The Agent Skills Directory

[SAFE]: The skill documents the official API and SDKs for the 'code.storage' platform, which is owned by the skill author 'pierrecomputer'. All referenced domains and repositories are consistent with this vendor.
[EXTERNAL_DOWNLOADS]: References official vendor-managed SDKs: '@pierre/storage' (NPM) and 'pierre-storage' (PyPI) for integration.
[DATA_EXFILTRATION]: Uses standard environment variables for authentication with the vendor's infrastructure. No unauthorized data collection or exfiltration patterns were found.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it can read and search the contents of external repositories.
Ingestion points: Reads file contents via GET /repos/file and POST /repos/grep as documented in SKILL.md.
Boundary markers: Absent. There are no instructions to the agent on how to isolate its logic from untrusted code content found in the repositories.
Capability inventory: Significant repository modification capabilities including commit creation (/repos/commit-pack), merging (/repos/merge), and credential management (/repos/git-credentials).
Sanitization: Absent. File content is processed raw as bytes or streamed without validation or sanitization.

code-storage